Journal of Cardiovascular Magnetic Resonance最新文献

Background: Cardiac diffusion tensor imaging (cDTI) non-invasively provides unique insights into cardiac microstructure. Current protocols require multiple breath-hold repetitions to achieve an adequate signal-to-noise ratio, resulting in lengthy scan times. The aim of this study was to develop a cDTI de-noising method that would enable the reduction of repetitions while preserving image quality.

Methods: We present a novel de-noising framework for cDTI acceleration centered on three fundamental advances as follows: (1) a paradigm shift from image-based to tensor-space de-noising that better preserves structural information, (2) an ensemble of Vision Transformer-based models specifically optimized for tensor processing through adversarial training, and (3) a sophisticated data augmentation strategy that maximizes training data utilization through dynamic repetition selection.

Results: Our approach reduces scan times by a factor of up to 4 while achieving a 20% reduction in cDTI maps errors over existing de-noising methods (fractional anisotropy errors 0.09 vs 0.07) and preserving anatomical features such as infarct characterization and transmural cardiomyocyte orientation patterns. Crucially, our proposed method succeeds in clinical cases where other algorithms previously failed.

Conclusion: This demonstrates substantial improvements in cDTI acquisition efficiency, achieving up to four-fold scan time reduction (3-5 breath-holds) while maintaining diagnostic accuracy across diverse cardiac pathologies.

Magnetic resonance imaging with gadolinium chelates has transformed the diagnostic practice of cardiovascular medicine. From anatomic and functional cardiac imaging to myocardial tissue composition, vascular imaging, and blood flow measurement, the clinical benefits of contrast-enhanced cardiovascular magnetic resonance (CMR) are well established. Since their introduction over 40 years ago, gadolinium chelates have had an excellent safety track record; however, concerns related to gadolinium retention have prompted in-depth consideration of potential risks and benefits in some patient groups. Recently, ferumoxytol has emerged as an off-label, versatile blood-pool contrast agent for CMR. Endowed with high r₁ and r₂ relaxivities, ferumoxytol is a clinically available intravenous iron supplement that was initially designed as a diagnostic agent and in 2025, was approved by the U.S. Food and Drug Administration for brain imaging. Because iron is vital for many biological processes, ferumoxytol spans both diagnostic and therapeutic dimensions. In this review, we summarize the attributes of ferumoxytol, highlight promising research directions, and illustrate several growing ferumoxytol-enhanced CMR applications. We conclude with a discussion of safety.

Background: Four-dimensional (4D) flow cardiovascular magnetic resonance (CMR) is a valuable technique for evaluating cardiovascular hemodynamics, but it requires cumbersome, offline preprocessing and regional segmentation prior to quantifications or visualizations.

Methods: The Framework for Image Reconstruction (FIRE) framework was used to integrate 4D flow processing tasks directly into the scanner reconstruction pipeline. The method builds containerized applications with standardized raw and image CMR data input/output in the open-source Magnetic Resonance Data format. In this study, deep learning models and algorithms from previous work (4D flow pre-processing, three-dimensional [3D] aorta segmentation, aorta velocity maps, quantification of aortic systolic peak velocities) were implemented in TensorFlow and executed within a containerized Python 3.6 environment on the magnetic resonance imaging (MRI) scanner, directly following the MRI data acquisition. All tasks were executed in-line using the MRI system's own computational resources. Analysis results were returned alongside the standard 4D flow CMR magnitude/phase images, available for review on-scanner console immediately after the CMR scan. In a study with 20 subjects (n = 10 patients with aortic disease, n = 10 healthy controls), FIRE performance was evaluated and compared to manual 4D flow analysis (reference standard).

Results: We successfully implemented on-scanner automated 4D flow hemodynamic analysis on a 1.5T MRI system. Total on-scanner computation time for 4D flow analysis was 220 ± 35 s. Dice scores between manual vs deep learning processing (eddy current static tissue selection: 0.84 ± 0.14; noise voxel detection: 0.92 ± 0.04; aortic 3D segmentation 0.92 ± 0.06) demonstrated good to excellent pipeline performance. Bland-Altman analysis revealed a small but significant bias (0.04 m/s, p = 0.01) for peak systolic velocities between manual and deep learning processing with good limits of agreement (-0.10, 0.18 m/s) and a mean relative difference of 4% (0.8/20).

Conclusion: An automated 4D flow processing workflow was successfully deployed for fully automated on-scanner hemodynamic analysis with good in-line vs human performance, indicating its potential for increased workflow efficiency.

Cardiac T1 and T2 mapping techniques are well-established methods for obtaining quantitative pixelwise representations of myocardial tissue properties. Mapping images are commonly evaluated quantitatively, and their resulting values play a crucial role in diagnosis and therapeutic decision-making in various cardiac pathologies. Despite the validated effectiveness of these techniques, both methodological and patient-specific confounders must be considered when applying them in clinical and research settings. Artifacts-erroneous features within the magnetic resonance image-can be misinterpreted as true anatomical structures or pathologies, potentially confounding quantitative analyses, conducted by both human readers and artificial intelligence algorithms. Artifacts can arise from sources such as patient motion, metal objects, hardware constraints, patient-specific scanner adjustments (e.g., flip-angle calibration), and processing errors, particularly within the complex environment of cardiac imaging. While artifact sources in other cardiovascular magnetic resonance sequences are well-documented, cardiac parametric mapping presents unique challenges due to its distinct image generation and quantitative assessment. This article provides an overview of artifacts encountered in cardiac T1 and T2 mapping, along with a concise explanation of their origins, aiming to raise awareness of their potential impact on clinical decision-making. Future developments, including sequences designed to mitigate mapping artifacts, are also briefly discussed. A strong interaction between scientists and clinicians is needed to overcome these challenges and maintain the reliability of quantification results.

Background: Late gadolinium enhancement (LGE) imaging remains the gold standard for assessing myocardial fibrosis and scarring, with left ventricular (LV) LGE presence and extent serving as a predictor of major adverse cardiac events (MACE). Despite its clinical significance, LGE-based LV scar quantification is not used routinely due to the labor-intensive manual segmentation and substantial inter-observer variability.

Methods: We developed ScarNet that synergistically combines a transformer-based encoder in medical segment anything model (MedSAM), which we fine-tuned with our dataset, and a convolution-based decoder in UNet with tailored attention blocks to automatically segment myocardial scar boundaries while maintaining anatomical context. This network was trained and fine-tuned on an existing database of 401 ischemic cardiomyopathy patients (4137 2D LGE images) with expert segmentation of myocardial and scar boundaries in LGE images, validated on 100 patients (1034 2D LGE images) during training, and tested on unseen set of 184 patients (1895 2D LGE images). Ablation studies were conducted to validate each architectural component's contribution.

Results: In 184 independent testing patients, ScarNet achieved accurate scar boundary segmentation (median DICE=0.912 [interquartile range (IQR): 0.863-0.944], concordance correlation coefficient [CCC]=0.963), significantly outperforming both MedSAM (median DICE=0.046 [IQR: 0.043-0.047], CCC=0.018) and nnU-Net (median DICE=0.638 [IQR: 0.604-0.661], CCC=0.734). For scar volume quantification, ScarNet demonstrated excellent agreement with manual analysis (CCC=0.995, percent bias=-0.63%, CoV=4.3%) compared to MedSAM (CCC=0.002, percent bias=-13.31%, CoV=130.3%) and nnU-Net (CCC=0.910, percent bias=-2.46%, CoV=20.3%). Similar trends were observed in the Monte Carlo simulations with noise perturbations. The overall accuracy was highest for ScarNet (sensitivity=95.3% (163/171); specificity=92.3% (12/13)), followed by nnU-Net (sensitivity=74.9% (128/171); specificity=69.2% (9/13)) and MedSAM (sensitivity=15.2% (26/171); specificity=92.3% (12/13)).

Conclusion: ScarNet outperformed MedSAM and nnU-Net for predicting myocardial and scar boundaries in LGE images of patients with ischemic cardiomyopathy. The Monte Carlo simulations demonstrated that ScarNet is less sensitive to noise perturbations than other tested networks.

Background: Exercise electrocardiogram (ECG) remains widely performed in the assessment of patients with suspected cardiac chest pain. We aimed to assess the comparative diagnostic and prognostic yield of exercise ECG, single photon emission computed tomography (SPECT), and cardiovascular magnetic resonance (CMR), in a large prospective patient population.

Methods: Patients recruited to Clinical Evaluation of MAgnetic Resonance in Coronary heart disease (CE-MARC) who had exercise ECG were included and followed up to a median (interquartile range) of 6.3 (0.1, 6.8) years. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the curve (AUC) for diagnostic accuracy were derived and hazard ratios of major adverse cardiovascular events (MACE) for prognostic significance were calculated.

Results: Of 752 patients in the CE-MARC trial, 580 had exercise ECG and invasive coronary angiography, of which 503 also had SPECT and CMR. At follow-up, a total of 91 (15.7%) patients experienced MACE. Using invasive angiography as the reference test, the sensitivity, specificity, PPV, and NPV (95% confidence interval) of exercise ECG were 68.3 (61.9, 74.0), 72.5 (67.6, 76.9), 61.0 (54.8, 66.8), 78.4 (73.7, 82.5). Exercise ECG was significantly less sensitive than CMR and less specific than both CMR and SPECT. A positive exercise ECG result was not predictive of MACE at follow-up (Hazard ratio 1.14 [0.75, 1.72], p = 0.53). CMR had both a greater diagnostic and prognostic yield than exercise ECG, SPECT, and their combination. Sequential CMR following inconclusive exercise ECG was comparable to CMR alone as the first-line test.

Conclusion: In patients with suspected angina, CMR alone as the first-line test was more sensitive and prognostically accurate than exercise ECG, SPECT, or sequential combination of both tests.

Background: Magnetic resonance imaging (MRI) with displacement encoding with stimulated echoes (DENSE) is well recognized for accurate and precise quantification of myocardial displacement and strain, but its reproducibility before and after contrast injection has not been investigated. Gadolinium is the most widely used contrast agent. Ferumoxytol is increasingly used off-label in specific patient groups. We aim to assess the reproducibility of cine DENSE MRI to measure global and segmental circumferential myocardial strain (E_CC) before and after contrast injection for gadolinium and ferumoxytol, respectively.

Methods: All imaging was conducted using 3T scanners. In 11 patients with cardiac disease, breath-hold two-dimensional cine DENSE was acquired in a mid-ventricular short-axis slice before and following the injection of gadolinium (0.1 mmol/kg). A separate cohort of 11 subjects (5 healthy subjects and 6 patients with ischemic heart disease) received 3 incremental doses of ferumoxytol: 0.125, 1.875, and 2.0 mg/kg (to a cumulative dose of 4.0 mg/kg). The same DENSE acquisition was performed before and after each incremental dose. Post-processing generated left ventricular (LV) displacement and E_CC maps, and strain-time curves. Global and segmental E_CC in six mid-level short-axis LV segments were compared. Signal-to-noise (SNR) was evaluated on the magnitude images throughout the cardiac cycle in the myocardium, liver, and back muscle, respectively. A Bayesian analysis was performed to test results with region of practical equivalence (ROPE) at ±5 for SNR and ±0.02 for E_CC (p < 0.05 as significant).

Results: Based on the percentage within the ROPE and the corresponding p-values, global E_CC exhibited excellent practical equivalence under pre- and post-contrast conditions for gadolinium (p = 0.413) and ferumoxytol (p ≥ 0.161). Segmental E_CC reproducibility was consistently high across all comparative analyses, with at least 87.02% falling within the ROPE. Gadolinium administration significantly improved SNR in all tissues during the early systolic phases (1-5, p ≤ 0.021). Ferumoxytol resulted in a reduction in liver SNR during diastolic phases (10-20, p ≤ 0.011) and a significant increase in myocardium SNR during systolic phases (1-5, p ≤ 0.034).

Conclusion: Good reproducibility of global and segmental E_CC measurements using cine DENSE before and after contrast injection is achievable at 3T.

Background: False-negative cardiovascular magnetic resonance (CMR) perfusion results may arise from inadequate stress responses, even when patients exhibit an adequate clinical or heart-rate response to adenosine. This study aimed to explore the ability of qualitative and quantitative splenic switch-off (SSO) markers to differentiate true-negative from potentially false-negative adenosine stress-perfusion CMR findings in a cohort where fractional flow reserve (FFR) was used to adjudicate lesion significance.

Methods: Patients with known or suspected coronary artery disease (CAD) from five centers underwent three-dimensional (3D) adenosine stress perfusion CMR and coronary angiography with FFR. SSO was assessed qualitatively using both standard stress-to-rest (SSO) and a stress-only (SSO_stress) approach. In addition, quantitative signal intensity (SI) ratios were assessed, including the splenic stress-to-rest SI-ratio (SI_stress/rest) and the spleen-to-myocardium SI ratio at stress (SI_{spleen/myocarcium}). The diagnostic accuracy of these measures was evaluated using cross-validated area under the curve (cvAUC) analysis.

Results: Among 179 patients (mean age 63 ± 10 years; 130 male), SSO prevalence was 73% (130/179) and was significantly more frequent in true-negative than false-negative CMR cases (80.6% [54/67] vs 36.8% [7/19], p < 0.001). SSO_stress showed moderate agreement (κ = 0.60) and robust diagnostic performance (AUC 0.80), as compared to SSO. Splenic SI_stress/rest and SI_{spleen/myocarcium} at stress demonstrated high predictive accuracy for visual SSO, with cvAUCs of 0.94 (95% CI: 0.90-0.96) and 0.90 (95% CI: 0.86-0.95), respectively. The positive likelihood ratio of SSO for true-negative CMR was 1.70, while the negative likelihood ratio was 0.24. Qualitative and quantitative splenic-switch off metrics classified 77%-80% (66-69/86) of negative CMR cases correctly as true- or potentially false-negatives, with sensitivities ranging from 81.4% to 91.2%. Clinically applicable cut-offs for differentiating true- and false-negative studies with splenic SI_stress/rest and SI_{spleen/myocarcium} at stress were identified as ≤0.32 and ≤0.38, respectively.

Conclusion: In a multicenter cohort using FFR-adjudicated reference for lesion severity, qualitative SSO and quantitative SI metrics were associated with myocardial stress adequacy and these markers may improve the interpretation of negative stress-perfusion CMR studies.

Background: Normal reference ranges in cardiovascular imaging studies are typically established as the mean value plus and minus twice the standard deviation (SD) of a healthy reference cohort ("2 SD-method"). Although widely used for cardiac magnetic resonance (CMR), this approach has not been previously validated. The purpose of this study was to use longitudinal cohort data to assess the clinical predictive validity of normal reference values for cardiac CMR.

Methods: Normal reference ranges for left- and right ventricular (LV and RV) CMR parameters were derived from baseline exam data of 1518 participants (age 45-84years) in the Multi-Ethnic Study of Atherosclerosis (MESA) study without known CV disease and without established CV risk factors. Cut-off values at 1 and 2 SDs were obtained for the following LV and RV parameters indexed to body surface area: end-diastolic volume (LVEDVi, RVEDVi), end-systolic volume (LVESVi, RVESVi), mass (LVMi, RVMi), as well as for LVED diameter (LVEDD), LVED wall thickness, and ejection fraction (LVEF, RVEF). The relationship of reference values to CV events was then evaluated in the entire MESA cohort with CMR data (n=4915), including individuals with CV risk factors at the baseline exam. Cox proportional hazard models were calculated for major adverse and all CV events (MACE and ACE, respectively) at 5 and 10 years of follow-up.

Results: At 5 years of follow-up, LVEDVi, LVESVi, and LVEF beyond the 2SD-threshold of the mean reference values were predictors of MACE and ACE in men and women (HR 2.1-4.3; P<.001-.029). In men, LVMi and LVED wall thickness above the 1 SD-threshold were associated with CV events (HR 1.6-2.1; P<.001-.002). For women, LVED wall thickness above the 1 SD-threshold significantly increased risk of adverse events (HR 1.6-2.3; P.034-.002) while LVMi was associated with events only for values above the 2SD-threshold (HR 2.7-4.1; P<.001). Notably, LVEDD, RVMi, RVESVi, and RVEF were not associated with CV events in men or women. CV events over 10 years showed similar trends.

Conclusion: Our results support the clinical relevance of CMR normal reference ranges for LV parameters. Most LV CMR parameters beyond the normal reference range (2SD-threshold) were associated with elevated CV risk at 5 and 10 years. Elevated LVEDDi, RVMi, RVESVi, and RVEF, however, were not associated with CV events.

Background: Despite its potential to improve the assessment of cardiovascular diseases, four-dimensional (4D) flow cardiovascular magnetic resonance (CMR) is hampered by long scan times. 4D flow CMR is conventionally acquired with three motion encodings and one reference encoding, as the three-dimensional velocity data are obtained by subtracting the phase of the reference from the phase of the motion encodings. In this study, we aim to use deep learning to predict the reference encoding from the three motion encodings for cardiovascular 4D flow.

Methods: A U-Net was trained with adversarial learning (U-Net_ADV) and with a velocity frequency-weighted loss function (U-Net_VEL) to predict the reference encoding from the three motion encodings obtained with a non-symmetric velocity-encoding scheme. Whole-heart 4D flow datasets from 126 patients with different types of cardiomyopathies were retrospectively included. The models were trained on 113 patients with a 5-fold cross-validation, and tested on 13 patients. Flow volumes in the aorta and pulmonary artery, mean and maximum velocity, total and maximum turbulent kinetic energy at peak systole in the cardiac chambers and main vessels were assessed.

Results: Three-dimensional velocity data reconstructed with the reference encoding predicted by deep learning agreed well with the velocities obtained with the reference encoding acquired at the scanner for both models. U-Net_ADV performed more consistently throughout the cardiac cycle and across the test subjects, while U-Net_VEL performed better for systolic velocities. Comprehensively, the largest error for flow volumes, maximum and mean velocities was -6.031% for maximum velocities in the right ventricle for the U-Net_ADV, and -6.92% for mean velocities in the right ventricle for U-Net_VEL. For total turbulent kinetic energy, the highest errors were in the left ventricle (-77.17%) for the U-Net_ADV, and in the right ventricle (24.96%) for the U-Net_VEL, while for maximum turbulent kinetic energy were in the pulmonary artery for both models, with a value of -15.5% for U-Net_ADV and 15.38% for the U-Net_VEL.

Conclusion: Deep learning-enabled referenceless 4D flow CMR permits velocities and flow volumes quantification comparable to conventional 4D flow. Omitting the reference encoding reduces the amount of acquired data by 25%, thus allowing shorter scan times or improved resolution, which is valuable for utilization in the clinical routine.